{
  "service": "ChronoSure",
  "available": true,
  "generated_at": "2026-08-04T16:18:10.735Z",
  "service_version": "0.1.0",
  "policy_version": "2026-08-03.1",
  "data_versions": {
    "tzdb": "2026c",
    "holidays": "2026.08.03",
    "temporal_source": "polyfill",
    "tzdb_source": "moment-timezone",
    "host_icu_tz": "2026b"
  },
  "corpus_size": 162,
  "methodology": "Cases are generated from the pinned tzdb's own transition table wherever a rule can produce them, so the expected offsets are read out of the dataset rather than typed by hand — an earlier hand-written truth table was wrong on 3 of 21 cases. Tiers run 1 (clean conversions) to 5 (adversarial: two-hour jumps, host/pinned tzdb disagreements, substitute bank holidays). Tier 2 exists specifically to catch false positives: ordinary times wrongly flagged ambiguous. Every case runs the same resolver a paying caller reaches, with no HTTP or payment layer.",
  "metrics": {
    "pass_rate": 1,
    "passed": 162,
    "failed": 0,
    "tier_1_pass_rate": 1,
    "tier_1_cases": 24,
    "tier_2_pass_rate": 1,
    "tier_2_cases": 33,
    "tier_3_pass_rate": 1,
    "tier_3_cases": 56,
    "tier_4_pass_rate": 1,
    "tier_4_cases": 29,
    "tier_5_pass_rate": 1,
    "tier_5_cases": 20
  },
  "false_positives": 0,
  "false_negatives": 0,
  "latency_ms_p50": 0.265,
  "latency_ms_p95": 0.693,
  "limitations": [
    "The corpus is self-authored. It is not an independent benchmark and no third party has verified it.",
    "Holiday coverage is bounded and enumerated: 17 jurisdictions (AU-NSW, CA-BC, CA-ON, DE, FR, GB-ENG, GB-NIR, GB-SCT, IE, IL, IN, JP, NL, NZ, US, US-CA, US-NY) for 2026-2027 only. Jurisdictions whose dates are set by annual gazette announcement rather than a published rule are deliberately absent, and any calendar not on that list returns UNSUPPORTED instead of an empty holiday list.",
    "The IANA tzdb release is pinned to 2026c via moment-timezone. Results describe that release. A later release can legitimately disagree, and the host runtime's own tzdb (currently 2026b) already does for some zones — those responses carry a warning naming the zone and both releases.",
    "Latency figures are in-process resolver timings from the eval run. They exclude network, TLS, and payment verification, and are not a substitute for the published SLA.",
    "Pass rate is measured against the pinned tzdb, so it measures faithfulness to that dataset rather than agreement with any external clock."
  ],
  "disclaimer": "These figures are evidence from a versioned corpus, not a guarantee. Corpus composition and size are stated so the numbers can be interpreted and reproduced."
}